Estimating the fatality Burden of SARS-CoV2

Image Credit: Laoise Tarrant

Statistics on Covid-19 are a constant presence in our new normal. Jade Norton looks at novel metrics developed by Professor Barry Smyth and whether a change in how we measure Covid-19 affects our view of the pandemic.

Every night the government of Ireland releases new statistics on COVID-19 which are a culmination of individual reports from across the country. These reports are the basis of statistical analysis and allow statisticians to model the progression of the pandemic. New novel metrics have been developed by Professor Barry Smyth, of the School of Computer Science UCD, that compare the historical death rate with the current death rate to see the effect COVID-19 is having in the wider timeframe.

The mortality rate caused by COVID-19 is used as a daily statistic to measure how we, as a population, are adhering to lockdown rules. These statistics are representative and not presented relative to general mortality rates which makes interpretation of whether they are a minor or large deviation from the large, difficult to know. In his paper in The Association for Computing Machinery, Smyth has developed a pair of novel fatality metrics that use historical mortality rates to normalise the current statistics and provide a comparable statistic between countries.

The two metrics developed were “the COVID-19 burden” and the “COVID-19 rank”. The former is defined by an equation which multiplies the proportion of COVID-19 deaths since the start of the pandemic with the expected death per day that gives the total number of fatalities across a historical axis. The former looks at how the death rate caused by COVID-19 compares to the effect of COVID-19 to historical causes of death in countries across the globe. The relative rank of COVID-19 in Ireland was 0.17 as of November 2020, which made it the 5th most common cause of death under cardiovascular disease (9,600 deaths per year), cancers (9,500 deaths), dementia (2,700 deaths), respiratory diseases (2,225 deaths), and lower respiratory infections (1,371 deaths).

One of the main ways of comparing the successful suppression of COVID-19 is to look internationally. The number of cases per capita is the most common variable used to measure a country's relative success at suppressing the spread of the virus. The way that each country reports a death varies between countries and is either: by the date of death, or the date on which the death was reported. Some countries such as Sweden choose to report deaths on a weekly rolling basis, which has made their spikes average-out as a day is left incomplete until all the data has been collected. The death rate per capita in Sweden, which opted to not implement harsh lockdowns, is several times lower than several European countries that opted for strict lockdowns.

Smyth used data from 174 countries to calculate both metrics, allowing an international comparison to be performed. It was seen that the COVID-19 burden increased over the winter months, especially in Europe where the effects of the first and second wave curtailed efforts to fight the next wave. The guideline burden across countries represented a single month of annual deaths, and approximately 80% of countries have a burden less than this. The worst cases of COVID-19 have been seen in a subset of 12 countries in Central and South America, which have a significantly higher burden representing almost two full months of deaths. Ireland has an expected death rate of ~400 per 100k people and COVID-19 represents 40 of these deaths, which is just under 1 month of deaths or 6% of yearly death rates.

Smyth’s model found that the large scale of the outbreak in South America accounted for 14% of historical deaths making it the 7th highest cause of death. However, the outbreak has significantly differed across the globe, and across the African continent, COVID-19 deaths account for just 1% of historical deaths. There has been speculation that this may be to do with the younger population present - the median age is 20 in Africa whilst it is 39 in much harder hit locations such as in the US.

The population spread has a larger impact on mortality rates as countries with older populations are far more susceptible to the virus. The historical deaths are not as consistent though, as with an ageing population the deaths have slowed over the last century due to increased access to healthcare and improved services.

The importance of developing statistics that accurately reflect the effect of COVID-19 on health is essential to understanding and building ways of living with it in society. Looking at the impact of COVID-19 in a wider context is important, as is looking at it with a combination of statistical and epidemiological expertise. This allows the healthcare system better capabilities to deal with the increased pressure and can help in the process of developing better strategies and more focussed attempts at balancing public health and economic stability. Real-time data is important in this effort, but it is essential that the data given to the public is correct and that the meaning behind it is clear.

Metrics such as this will be of more use in the further as they will allow governments to come back and “assess how countries responded and performed during the pandemic”, The COVID-19 burden metric is useful here as it allows the state of the country to be measured over an extended period, not just in the present. Smyth states that “Fortunately for us, the government’s modelling group is well suited to the task at hand and brings together both disciplines so I’m pretty comfortable with, and have confidence in, the current state of our statistical modelling in Ireland”.

Additionally, Smyth has been working on other metrics to estimate the ‘exposure risk’ of COVID-19, which is “ the likelihood that we will come into contact with an infected individual and it is based on an estimate of the number of undetected infected individuals at a given point in time”. And he believes that this could possibly be more useful than the conventional metrics as it would help people to calibrate their behaviour in real-time.